The Regional Factor in Contemporary Ukrainian Politics: Scale, Place, Space or Bogus Effect? 1 John O’Loughlin Institute of Behavioral Science and Department of Geography University of Colorado Campus Box 487 Boulder, CO. 80309-0487 Email: [email protected]1 This research was supported by a grant from the Geography and Regional Science Program of the National Science Foundation. Special thanks are due to my research assistant, Altinay Kuchukeeva, for her care in the retrieval of the Ukrainian electoral data and her cartographic diligence. As always, Jim Robb of the Geography Department’ s Cartographic Services Laboratory went far beyond expectations in preparing the figures for publication.
54
Embed
The Regional Factor in Contemporary Ukrainian Politics: Scale
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
1 This research was supported by a grant from the Geography and Regional Science Program of the National Science Foundation. Special thanks are due to my research assistant, Altinay Kuchukeeva, for her care in the retrieval of the Ukrainian electoral data and her cartographic diligence. As always, Jim Robb of the Geography Department’s Cartographic Services Laboratory went far beyond expectations in preparing the figures for publication.
1
Abstract
Post-independence Ukraine continues to be the subject of intense interest about its regional political divisions
and ethnic/language tensions. The debate between the school arguing that regional cleavages are ebbing and
thatwhich holds that Ukraine has not yet become a political community, is also fundamentally a geographic
question regarding scale and place. Using two measures of political preferences, votes in the 1999
Presidential runoff election and the political attitudes expressed in 1992 and 1996 Eurobarometer surveys, the
regional effect in Ukraine is shown to be complicated by the nature of the political question and by local
disparities from regional trends. New methods of analysis and graphical display of statistical results clarify
these complications and challenge both schools of researchers to pay heed to issues of measurement,
This asymmetry of language and nationality has led to the weakest sense of national identity among the
Russophone Ukrainians, who are mostly “Ukrainian” in political terms, and mostly “Russian” in terms of
culture (Ryabchouk, 1999). With respect to geographic distribution, the west of the country has a small
Russian minority with a strong preponderance of Ukrainian-speaking Ukrainians, the center of the country
(including Kyiv, the capital) has a large Ukrainian majority but a mixed Russian-Ukrainian speaking
population; the south has a Ukrainian majority but a Russian-majority (over 60%) in Crimea and a mixture of
languages; and the east has close to a 50-50 ratio of the two nationalities but a Russian-speaking majority
(Janmaat, 1999). In 1989, the last Soviet census showed an Ukrainophone Ukrainian percentage ratio in the
regional capitals equal to 76.6% in L’viv, 60.0% in Kyiv, 29.6% in Odessa, and 18.3% in Donets’k (Arel and
Khmel’ko, 1996). By contrast, the Russophone Ukrainian percentages ranged from 16.1% in L’viv, 15.4% in
Kyiv, 19.3% in Odessa, and 21.1% in Donets’k. Though many Russians claim to speak Ukrainian fluently, in
practice almost all members of this ethnic group chose Russian as the language of the home (Janmaat, 1999;
Lieven, 1999). Official language policies since 1991 to promote Ukrainian as the national language are
making inroads in the Russian populations in L’viv and Kyiv, especially in the areas of schooling and state
institutions, but are encountering resistance in Odessa and Donets’k (Janmaat, 1999). The high rate of inter-
marriage between Ukrainians and Russians and the presence of children in these families, especially in the
center and east of the country, do not allow a simple equation of region and nationality. Multiple identities
characterize a large proportion of the population; a strong feeling of attachment to the locality is added to the
cultural-linguistic, national and state identities (Pirie, 1996; Lieven 1999).
For social and political geographers, the influence of regional loyalties and a strong sense of locale
are well documented for contexts in Western Europe and North America. The hollowing-out of the
5
European states has occurred in the face of the growth of the European Union at the super-national scale and
the (often-reluctant) recognition by national governments of the demands for more autonomy at the regional
level (Delamaide, 1995). Debates between political geographers, like Agnew (1996), and political scientists
highlighted the issue of whether regional peculiarities are ebbing in the face of nationalizing politics in the
Western states. According to the classic political science model, during the early stages of democratic politics
(either during the early years of mass suffrage or consequent on independence), parties and political
movements will tend to have strong regional components that wither as the parties begin to broaden their
appeal beyond traditional strongholds and adopt national platforms. The hypothesized end-results of this
process are political patterns that do not vary from region to region, but instead are explained by
compositional factors (class, gender, age, education, etc). The United States is often viewed as a good
example of this nationalization thesis. Hinich, Khmelko and Ordeshook (1999, 182) make the comparison
explicit by arguing that the apparent regional diversity in Ukrainian politics today is probably no greater than
the early post-colonial years of the United States.
Ukraine is officially in the process of building a “civic nation’, one whose ideals are not ethnic-based
but that transcend national interests in the goal of uniting all residents as part of the Ukrainian nation. While
the language policies emanating from parliament and governments in Kyiv have caused some concern in the
Russian-majority cities, the slow pace of language policy change in the educational and governmental spheres,
as well as the absence of any rules that target minorities for special membership qualifications, have eased
minority concerns about the nature of the post-1991 Ukrainian independent state. Janmaat (1999) finds
regional differences in language retention policies – Russians opt for language retention in Donets’k and
Odessa, for language integration in Kyiv, and depending on family situation for assimilation (Russians in
mixed couples) or retention (Russians in purely Russian families) in L’viv; Janmaat thinks that fear of cultural
loss may have prompted the response of Russians in L’viv. It is important to keep in mind that Ukrainians
prefer Russians more than any other nationality (Miller, White and Heywood, 1998, 446).
In a sense, the relative lack of nationalist push factors from Kyiv has not created a regionalist
backlash in the minority areas, as was the case in neighboring Moldova after independence in 1991. But it is
6
still an open question if Ukraine is coming together in a political community, becoming more regionalized, or
essentially remains in a state of little change. In opposition to the Lieven argument for the tolerance of the
Ukrainian state stands the position that to speak of “Ukrainian” as some sort of unitary or homogenous
identity is to assign to a population a position that does not (yet) exist. Zimmerman (1998) believes that the
existence of an Ukrainian political community (when citizens in a territory share a range of values and
perceptions that distinguish them attitudinally) is still an open question. While regional cleavages are
prominent in the Ukrainian social, cultural, linguistic and political landscapes, they show no evident signs of
becoming mobilized into regional separatist or nationalist movements. Though there is significant differences
between the 3 regions (west, center and south/east) on key questions like “Ukraine and Russia must be
absolutely separate countries”, the difference is less than Ukrainians assumed. Asked to estimate strength of
feelings on this question, they guessed right for eastern Ukraine but over-estimated the support of Ukrainian
independence in the west of the country.
In quantitative geography, the term “regional effect” can have multiple meanings and can be caused
by two different, but related, processes. We thus need to separate “spatial dependence” (the contagion
effects of adjoining or neighboring spatial units) from “spatial heterogeneity” (usually considered to be
equivalent to the regional effect). Clusters of similar political patterns can be produced from spatial
dependence that shades into spatial heterogeneity. In order to test the validity of the claims about regional
effect, we need to use the smallest units available to see if the regional effect is a function of scale. In the case
of Ukraine, it is plausible that one could mistake clustering of the 26 oblasts of Ukraine for evidence of a
regional effect but examination of cities and rayoni (rural districts) could reveal intra-oblast discrepancies that
undermine any notion of homogenous oblast. Local and oblast-level patterns could show different trends
and support contradictory hypotheses regarding the significance of the geographic effect in Ukraine.
Unfortunately, most work to date on the political geography of Ukraine has been at the oblast level or even at
the macro-regional scale. Examination of the political data at a variety of scales should help to remove some
of the confusion about the significance of the regional effect in post-independence Ukraine. The strong
7
emphasis in this paper on methods of visual display can be helpful in defining the nature of the level of
aggregation and in turn, the displays can clarify the most promising paths to further analysis.
Regions in Ukraine
In the former Communist states of Eastern Europe and the Soviet Union, the legacies of regional histories lie
heavy on the contemporary political landscapes. In the territories of the former Austro-Hungarian Empire,
regional historical memories are being rediscovered and re-created in attempts to highlight traditions that
allow regional interest groups to certify their European credentials and to attract tourists and investment from
the West (Bialasiewicz and O’Loughlin, 2001). In countries such as Russia (Kolossov, 1993) and Poland
(Zarycki, 2000), regional divisions of the past are easily visible in the contemporary political maps. Ukraine
has a complex regional mosaic developing out of its centuries of division between the Russian, Austro-
Hungarian and Turkish empires and the imprint of these legacies are still visible in the electoral maps (Birch,
2000). The contemporary boundaries of the state date only from 1945, and it is especially significant that the
region of Galicia in the far west, distinctive on all Ukrainian political and economic maps (Craumer and Clem,
1999; Holdar, 1995; O’Loughlin and Bell, 1999; Wilson, 1997; Wilson and Birch, 1999), was a Polish territory
until the end of World War II. While Ukraine can be dichotomously divided at first glance along the Dnipro
river into east and west, a more nuanced political perspective would separate the state into four (east, west,
center and south) or five (further division of the south region into Crimea and the rest) macro-territories.
The regional patterns in political preferences come through clearly in relation to Ukrainian foreign
policy with two broad preferences apparent which can be usefully, though perhaps simplistically, represented
as those who would prefer Ukraine to adopt a “Slavic choice” and those who support a “European choice”
(Light, White and Löwenhardt, 2000, 82-83). Communists and other left-wing groups favor the restoration
of the Soviet Union or at the least, a Slavic confederation with Russia and Belarus; they live predominantly in
the east, south and Crimea. Those who favor a European choice tend to be centrists by political conviction,
staunch defenders of Ukrainian sovereignty, in particular its independence from Russia, though they
8
understand that Ukraine must have good relations with Russia; they tend to predominate in the center,
including Kyiv, and the west. Shulman’s (1999) survey of approximately 1000 members of the elites in L’viv
and Donets’k confirms the powerful pull of external linkages to Russia in the east and the degree of alienation
of elites in the west towards the Donbas. On the basis of these answers, he argues that this “asymmetrical
international integration” weakens the state and undermines national unity in Ukraine.
There is no doubt that a strong geographic pattern exists in other aspects of Ukrainian political life.
In general, from west to east, there is a fairly regular distance-decay in support for the nationalist parties in
the Verkhovna Rada (Supreme Parliament) elections in March 1998 (Wilson and Birch, 1999). In 1994,
President Leonid Kuchma received his strongest support in the east and south in running against the
incumbent President Leonid Kravchuk (Holdar, 1995). Crimea is the most “Russified” region of the country
and other southern regions also behave politically more like the east. Kyiv is split between nationalists and
Communists and the center is generally becoming more nationalist over time (Craumer and Clem, 1999). In
economic terms, the east is more industrialized than the rest of the country and in general, has higher
incomes and a more urbanized population.
In the numerous surveys of Ukrainians taken since 1991, one of the most remarkable features is the
east-west split in perspectives on the future prospects for the Ukrainian state. Respondents in the west are
significantly more optimistic than other Ukrainians and more supportive of the attempts of the Ukrainian
regimes since 1991 to reduce economic dependence on Russia and promote political and economic ties to the
West. Even simple measures of civic engagement, such as membership of clubs and organizations, also show
a strong east-west gradient, further evidence of the acceptance and optimism that people in the western part
of Ukraine hold for the new state and society (O’Loughlin and Bell, 1999). Attempting to explain these
differences, Åberg (2000) uses another survey of respondents in L’viv and Donets’k to argue for the
persistence of non-communtarian social capital in the post-independence Ukrainian west as a device for
practical problem solving in a time of economic difficulties. Residents of L’viv are much more likely to join
organizations, to sign petitions, to contact government officials and to participate in demonstrations that
residents of Donets’k.
9
In his dissertation, a study of political life in L’viv (west) and Donets’k (east), Clem (1995) compares
the regional differences that exist on all the types of measures of political institutionalization. He concludes
that in western Ukraine, reform-minded party activists successfully co-opted local pre-independence power
structures but in eastern Ukraine, leftist parties used preexisting ties to local government and other resource
providers to maintain their dominance in the region. In a parallel survey, Shulman (1998) finds that elites in
L’viv are suspicious of the ethnic and national ties between Russians in the Donbas and fellow Russians
across the border, believing that these ties tug on Russian loyalties in the Donbas: elites in Donets’k strongly
reject this claim. To prevent against the possibility of further integration with Russia and a loss of political
and economic independence for Ukraine, nationalists in the west want a unitary territorial-administrative
structure, while for the elite residents of the Donbas, a more federal structure that would allow their region
more autonomy including the chance to intensify cross-border relations with neighboring Russian regions
(Kolossov and O’Loughlin, 1998).
Underlying most explanations of the regional patterns, both statistical and historical, is the linguistic
distribution of Russian and Ukrainian-speakers and the associated, but incomplete, correlation with ethnic
Ukrainians and Russians. One of the major difficulties with the national identity literature that has emerged
in Europe in the past few decades is the assumption that individual members of a nation will hold fast to a
single (national) identity and that over time, other members born into the group or entering through marriage
or immigration will assimilate to this single identity. Recent research has challenged this assumption and has
shown that individuals can have multiple identities and place attachments. In the former Soviet Union, the
state promoted a Soviet identity that was supposed to supersede ethnic and national orientations but in
practice, tended to become layered on top of the local and republic attachments (Kaiser 1994). In Ukraine,
layering in a kind of “matrioshka doll” fashion became common especially because of close relations between
the two language communities and the high rate of inter-marriage across ethnic lines (Taras, 1993). The
Ukrainian scholar, Taras Kuzio (1997, 1998) assumes that Ukraine is becoming a modern state as national
identity is both promoted by the state apparatus and adopted by most residents. For those residents who
reject these practices, he labels the self-professed Soviet identity of some eastern Ukrainians as ”pre-modern
10
or transitional” claiming that “national identities are indispensable for political reform because only in nation-
states have democracies been traditionally created” (Kuzio, 1998, 144), a position challenged by Flynn (2000).
She particularly disputes Kuzio’s claims that civic society is weakest in south-eastern Ukraine because national
identity is weakest there.
The dispute between Kuzio and Flynn reflects a bigger debate between promoters of unitary national
identities for the newly-independent states of formerly-Communist Europe and those who insist that this
centralization will inevitably lead to “backlash nationalisms” as minorities become mobilized in the face of
nationally constituted majority practices. What distinguishes Ukraine from this conflict scenario is the
common belief across the majority-minority divide that all sit in the same leaking economic boat, though in
general residents in the western region remain more optimistic about their economic and political futures,
expecting a strengthening of ties to the West. Russian identity in the east and south of the country is much
more tied to the Russian heritage and people, and not to the idea of unification with the state of Russia
(Lieven, 1999, 141). Thus, to speak of a bifurcated state or ethnic mobilization in Ukraine is certainly
premature. The fact that no party or group has developed since 1991 to represent all the strongly-Russian
oblasts, despite the severe decline of the Soviet planned regional economy, can be seen as further evidence of
the lack of identity based on language or ethnicity (Lieven, 1999). Instead, fierce competition at the regional
and national levels between cadres of political and economic elites, especially those from Donets’k and
Dnipropetrovs’k, has characterized political life in the east of Ukraine.
It has become evident from surveys of elites (Shulman, 1998 and 1999; Clem, 1995) and the public
(Arel and Kmelko, 1996; Hesli, Reisinger and Miller, 1998) that ”the Russian question”, consisting of the dual
elements of the nature of the relations of the Ukrainian and Russian states, as well as the cultural and political
expression of the “Russophone” population in Ukraine, overrides other polarizing issues in post-
independence Ukraine. As noted by Hesli et al (1998), the shared misery of a declining standard of living
since 1991 helps to unite Ukrainians across ethnic, linguistic and regional lines. However, there are some
sizeable differences by region and ethnic/linguistic groups on the question of the best strategy of dealing with
the shrinking economy. Residents of western Ukraine, especially the three Galician oblasts of Ivano-
11
Frankivs’k, Ternopil’ and L’viv, show significantly more support for the privatization strategy enounced,
though not effectively pursued by successive Ukrainian governments, than residents of other regions,
especially in the Donbas industrial agglomeration (Arel and Kmelko (1996). Part of the explanation for the
differences in attitudes towards privatization could be attributed to the fact that privatization of the large
industrial enterprises (coal mines and steel mills, for example) is unlikely to be successful, while the chances
of success of the small, often agricultural, enterprises of the West look more promising.
The “Russian question” touches on both internal and external relations in Ukraine, making the issue
doubly sensitive. The sensitivity can recently be judged by consideration by the state of instituting Ukrainian
as the only language for official state business after the Constitutional Court ruled that all state officials
should know and use Ukrainian; this proposal generated a major backlash in the Russian-speaking area.
Local officials in L’viv (Galicia) went farther by limiting the use of Russian in public places, including popular
music, and business (Kuzio, 2000). Earlier surveys between 1991 and 1994, reported in Arel and Kmelko
(1996), clearly indicate that the most sensitive issue in Ukraine was the status of the Ukrainian state vis-à-vis
Russia. Enveloped in that sensitivity is the worry for some Ukrainian created by the “pro-Russian” attitudes
of a large segment of the population, especially in the east. Only the composite index, “pro-Russian” elicits a
significant territorial polarization in Ukraine and Arel and Kmelko (1996, 88) conclude that “if (their italics)
Ukrainian politics is territorially split at a given moment, the sources of the split are to be predominantly
ascribed to the clashing attitudes of the electorate over the Russian question”.
Layered on the cultural and linguistic identities are local identities, circumscribed by locality or oblast.
According to Lieven (1998, 79), “Many Ukrainians could still be plausibly described as tuteshni – that is to say,
people whose primary identification is with their locality rather than with their state or ‘nation.’” Ryabchouk
(1999) indicates that this local identity is strongest for Russophone Ukrainians (about 30% of the national
total), many of whom have a mixed and vague identity and who usually identify themselves in pre-modern
terms as ‘locals’ (”Odessans’, “Kyivans”, “Donbasians.” In the Donbas, the industrial heartland of Soviet
eastern Ukraine, the region’s economic raison d’etre has been damaged by the loss of centrally-planned markets
in other former Soviet republics. Economic dislocation, felt most severely by coal miners and their
12
communities, has been partially transformed into a stronger sense of betrayal by the Ukrainian state than is
felt in other (also economically depressed) regions. Many respondents in this region hark back nostalgically
to the banner years of the Soviet state and enounce a stronger Soviet identity than other regions of Ukraine
(or indeed, of many parts of the other Soviet republics) (Kolossov and O’Loughlin, 1998).
The summary of the literature on Ukrainian identity indicates that expectations of ethnic-based
conflict have been proven wrong by accommodations on all sides, including the state and dominant political
figures, since 1991. While the electoral maps of Ukraine seem to indicate a strong east-west divide, the same
geographic cleavages can be observed in other democratic states, such as Italy, the United Kingdom, and
Germany. The electoral geographic cleavage can be produced by many factors, especially the clustering of
compositional groups (classes, religions, ethnicities, urban populations, etc) differentially across the regions.
The unresolved question is whether the east-west divide remains visible when these factors are taken into
account. In preparing the statistical analysis of the influence of these compositional elements, we need to
remain attuned to the warning of Lieven (1998, 80) in summarizing the Ukrainian regions debate.
“(Commentators) have missed three important elements of Ukrainian political geography: the fact that
nationalist Galicia does not make up the whole of ‘western Ukraine’ and that its specific variant of
nationalism has very limited cultural and economic appeal outside its own region, the critical importance of
central Ukraine, and the divisions within the whole of the Russian-speaking area.” The analysis in this paper
specifically incorporates these three elements in the determination of the size and importance of the regional
factor in Ukrainian political life.
Statistically isolating the regional factor in Ukrainian politics: From the literature examined above, it is clear
that significant differences in political preferences, ethnic identification, national mobilization, regional
loyalties, and language use exist between west and east/south Ukraine. It is also evident that these elements
overlap to such an extent that it is difficult to isolate the relative importance of each. In the past 5 years,
numerous studies have used statistical methodologies to determine the relative importance of the regional
factor, controlling for other compositional characteristics (ethnicity, language use, age, education, etc).
13
Though the studies report regression-type results, they use different polls and the comparison of the
conclusions is thus not as conclusive as might be expected. In general, the conclusion is that the regional
factor exists independent of the compositional effects, though why it persists is not evident.
In their comparative surveys of post-Communist societies in eastern Europe and the former Soviet
Union, Miller, White and Heywood (1998) use broad regional divisions to see if there are any consistent
elements in the survey responses. While their tripartite regional division of Ukraine (east, west, and center)
does not include all oblasts, it is used as a sampling framework but, unfortunately, regional controls or
variables are not incorporated into their statistical models. It is noticeable that the fits of their statistical
models using only compositional variables are lower in Ukraine that the other countries surveyed (Russia,
Czech Republic, Hungary, Lithuania, and Slovakia). In a similar comparative study of social identities in
Russia, Ukraine and Lithuania between 1995 and 1997, Miller, Klobucar, Reisinger and Hesli (1998) found
that the explanatory regression model for Ukraine was significantly improved by dividing the sample into
respondents from west and east Ukraine. In east Ukraine, political orientation was strongly connected to class
identification, while in the west, ethnic identification was dominant, leading to the conclusion that in this
region, a strong national identity promotes democracy and opposition to communists. However, the authors
optimistically conclude that, in Ukraine, as in Russia and Lithuania, class interests are replacing ethnic
identification as the major factor behind the consolidation of state identification.
Studies of the regional factor in Ukraine have frequently used public opinion surveys of political
attitudes and aspirations for the future. While there is a hypothesized translation of these attitudes into
political party preferences, the correlation is not precise and it is expected that regional traditions and
historical patterns of political behavior will modify the national model. In a study comparing 1995 and 1997
survey data for Ukraine, Hesli, Reisinger and Miller (1998, 237) argue that “national integration is occurring in
the sense that the Ukrainian electorate is becoming less polarized over time, despite the existence of deep
historically-based cleavages in the society.” Like other studies (e.g. Arel and Kmelko, 1996), they use the
“Russian question” to isolate the level of regional and political party polarization. Not surprisingly, they find
that religion, language, nationality, region and party identification are all intertwined and that each makes a
14
contribution to the polarization of Ukrainian society. In a regression model, a comparison of the
standardized regression coefficients leads to the conclusion that region, measured by residence in the western
part of Ukraine, is the most important determinant of answers in 1995 on the “Russian question”, followed
by self-identification as “Ukrainian”: it is important to note that party affiliation (Communist, nationalist, etc)
was less important. By 1997, the party affiliation indicator had moved into second place behind region.
These models certainly offer no support for the ebbing of the regional factor in Ukraine; for that claim, the
authors rely on a logit model of voting choice for either Communist or nationalist candidates. Negative
orientations on the “Russian question” and use of the Ukrainian language offer the most important
explanations of the vote choice, with the regional variable (residence in west Ukraine) lagging behind. It
would appear that the conclusions of the study should concern the nature of the issue; for the “Russian issue”
debate, regional location is clearly still most important, while other questions require different explanations.
It is unsurprising why this regional difference on the “Russian question” should persist since the subject goes
to the heart of Ukrainian independence and separation from Russian dominance. The study, however, does
not mark the end of the search for an account of the regional factor in Ukraine.
Kubicek (2000) challenges the Hesli et al (1998) study’s conclusion about the decline of the regional
factor. Using time-series Eurobarometer data (same as this study), electoral results and voting in the Rada by
deputies, he argues for the persistence of regional divisions. However, a closer look at the statistical support
for Kubicek’s conclusion reveals a mixed picture. The regional variables (west, south, east, center, Crimea)
vary greatly in their significance according to the survey question and year (Kubicek, 2000, Table 3, p. 281).
Furthermore, the absence of any party and other political controls on the distribution of deputies’ voting in
the Rada exaggerates the level of their regional polarization. Using a series of dummy variables for language
groups, ethnicity, and regional location for a 1994 national survey on loyalty to the state of Ukraine,
Barrington (1997) comes to the same conclusion as Kubicek (2000), that region is not declining in
significance in post-independence Ukraine. However, once more, the study fails to account for the
independent effects of region, language and ethnicity, arguing that they are once again intertwined.
15
“Is the East-West divide in Ukraine so deep that voters from one geographic region see a different
political universe than do voters from the other?” is the direct question asked by Hinich, Khmelko and
Ordeshook (1999, 152), a question that lies at the heart of the many research papers on Ukrainian
regionalism. Using data from a large sample (2923 respondents) in early 1998, they asked respondents to rate
themselves on an ideological scale and they then examine scale positions using respondent characteristics.
The greatest variation in the “ideal points” (a self-identified position on two ideological scales) among
different groups is found for the 26 oblasts of Ukraine but when individual voters are mapped in ideological
space by region, there exists quite a bit of overlap in their ideological spaces. The authors conclude that this
overlap offers some room for optimism because it might allow a “centrist” party to appeal to this
compromise position, thus undermining the regional identification. So far, no national party has filled this
ideological vacuum in Ukraine and only the Communist party is a national party by virtue of its organization
across the geographic units of the country.
We can conclude from the mass of studies over the past decade that the Ukrainian political
community has not yet fully formed. While there is some evidence that the regional factor is becoming less
important for some issues, such as the preferences for the capitalist or communist economic model, it persists
strongly for issues surrounding the “Russian question.” Because parties tend to have regional bailiwicks,
measures of regional polarization that use party votes, party memberships, deputy behavior in parliament and
other “formal political” measures will tend to show greater levels of polarization of the electorate. Public
opinion surveys, by contrast, offer more evidence for compromise positions because the correlates of
preferences to ethnic, national, regional and especially compositional (age, education, gender, etc) variables
are relatively weak. In the empirical part of this paper, I will use both survey data and voting outcomes as
measures in the attempt to clarify further the nature and scope of the regional polarization of Ukraine.
Data and Methodologies
16
To filter out the different kinds of possible causes of the so-called “regional factor”, we need to employ
specific statistical methodologies that have been developed in Geography and Political Science in the past
decade. Unfortunately, the full suite of spatial techniques cannot be deployed since the data available for
analysis are limited, inconsistent across time and region, and do not include the usual variable mix that are
incorporated into census materials. Ukraine has not had a national census since the last Soviet census of 1989
and given the dramatic economic and population changes since then, use of these data to reflect
contemporary developments would be highly questionable.
Electoral returns for recent Ukrainian elections are the most reliable data available and also possess
the advantage of national coverage. Parliamentary elections were held in March 1998 and were the subject of
a recent paper by Craumer and Clem (1999) that indicated a strong regional factor in the oblast returns.
Dissection of this factor is difficult since there are no aggregate socio-economic correlates available at the
rayon or constituency level (225 in the country) and party formation is still evolving in the new democracy.
With the exception of the Communist party, parties do not have a national range, appeal or organization, and
further, they tend to be non-ideological, strongly focused on personalities, and unstable in membership and
loyalty. In the current (June 2000) Verkhovna Rada (parliament), 13 fractions were registered but 35 deputies
stated no fraction preference instead opting for a “regions” label (Laboratory F-4, 2000). Most of these
deputies are independents who have aligned themselves with local “parties of power” in the oblast capitals.
The 1998 Presidential election run-off in Ukraine offered a clear ideological choice to the electorate.
President Leonid Kuchma (originally a Russian-speaking missile factory manager from Dnipropetrovs’k) had
been elected with strong support from the east and south in the 1994 election, but over his five year term, he
espoused a moderate Ukrainian nationalist position, aligning himself with the west and center of the country.
In the first round of the Presidential contest, he won 36% against 12 opponents on a platform of continuing
the pro-West policies of his first term, of more privatization, of promoting Presidential authority in the face
of parliamentary opposition, of instituting more control of the government, and of changing the constitution
to encourage more stable government and fewer parties. His runoff opponent, Petro Symonenko from
Donets’k, won 22% of the first round total as head of the Ukrainian Communist party. His party had
17
emerged in the March 1998 parliamentary elections with 26% of the vote and formed the most coherent
opposition in the Rada to the policies of President Kuchma and his prime ministers. Symonenko barely
edged two other leftist candidate in the first round and therefore, allowed President Kuchma to portray the
run-off as a stark choice between an independent Ukraine and a return to a Soviet-style economy, society and
identity for Ukrainians. In the run-off on November 14, 1999, Kuchma received 56% to Symonenko’s 37%
percent on a national turnout of 79%. For the run-off, the votes for the candidates as well as other
constituency data (urban or rural district, turnout rate, valid votes) were used in the analyses. The sources of
the data are the official returns of the Ukrainian Election Commission available from International
Foundation for Election Study (www.ifes.kiev.ua). Because the run-off dispensed with any confusion
generated by multiple candidates, the aggregate statistics are more representative of voter preferences and
Ukrainian electoral divisions.
The second data set has been used widely in the study of the evolving political beliefs of the citizens
of the new democracies of Central and Eastern Europe (Haepfer, Milosinski and Wallace, 1999). Conducted
for the Eurobarometer by local polling firms, the yearly survey included Ukraine from 1991 until 1996; since
then, the surveys have been confined to those former Communist countries that have aspirations to join the
European Union. While many of the questions concern attitudes toward the European Union, standard
questions also asked for opinions on democracy, aspirations for the country, language used, ethnic
identification, relations with neighboring states (including Russia), the rate of privatization, as well as the usual
compositional questions (age, gender, education, income, subjective measure of the standard of living, and
regional location – 10 in the case of Ukraine). The large sample size in each country was designed to
produced a margin of error about 3%. In the case of Ukraine, the sample size was 1400 in 1992 and 1200 in
1996, the two years used in this analysis. While the Eurobarometer surveys allow some geographic analysis at
the macro-regional level (groups of 3-4 Ukrainian oblasts, termed the Northwest, West, Southwest, etc. in the
survey), they are not fine-grained enough to match to oblast-level aggregate results for elections and other
political expressions. In this study, I used the Eurobarometer surveys to construct logit political preference
models and then re-calculated the simulated regional mean preferences as well as the means for other specific
18
populations (Ukrainian-speakers in the west, 65 year old people, urban residents of the south, etc). By
comparing the expected regional means for the sub-groups using a simulation model, we can gauge the
relative significance of the regional factor. The null hypothesis is that the simulated means will not display
any significant differences across the four macro-regions (east, center, west, south). The Eurobarometer data
are available from the Inter-University Consortium for Political and Social Research (ICPSR) at the University
of Michigan (www.icpsr.umich.edu).
This article offers the first combined use of two statistical methodologies. Each offers a specific
advantage in answering the question posed in the sub-title of this paper. Is the regional factor in Ukraine the
product of some scale effects caused by the over-reliance on oblast-level data and where the use of data at a
finer spatial resolution (for constituencies and rayoni) does not support a regional explanation? Perhaps the
regional factor is a spatial artifact - that is, it is the manifestation of a clustering effect that cannot be
removed by the incorporation of more compositional variables. King (1996), in responding to the statement
of John Agnew (1996) about the meaning and expression of contextual effects in politics, argued that
“geography should not count”, meaning that proper statistical analysis combined with appropriate data and a
good theory should be able to account for any spatial variation in the political phenomenon under study. But
as is well demonstrated in dozens of statistical analyses of elections, significant spatial clustering of error
terms in regression equations cannot be easily explained away (O’Loughlin, 2001). We can call these
remaining clusters of significant residuals the “effects of space”. When we can identify a geographic factor
that correlates well with the cluster, such in a region with a distinctive history and identifiable political or
cultural profile, we can call this factor the “place effect”. Here I follow Tuan’s (1977) place-space distinction,
an approach continued recently by Taylor (1999). If the “regional effect” disappears in the face of careful
statistical controls and appropriate methodologies, I will refer to this “effect” as bogus.
It has been the norm in electoral geography to calibrate regression models that use compositional
variables as independent predictors of the dependent variable, the vote percentage for a particular candidate
or party. To avoid falling into the trap of the “ecological fallacy”, geographers and other social scientists
19
have resorted to such statements as “Counties with large percentages of African-Americans supported the
Democratic candidate.” But what geographers typically cannot do is infer the ratio of the population of
interest (say, African-Americans in Alabama) that voted for the Democratic candidate nor can we provide
small-scale estimates (say, for precincts or counties) of the ratio using aggregate data available from the census
and from election commissions. While surveys of individuals are widely used by political scientists,
geographers tend to rely more on aggregate data that can be mapped and interpreted. A recent development
by the political methodologist, Gary King (1997), has combined two existing estimation methods of inferring
individual votes from aggregate data. Kings’ method allows not only estimates of individual behavior but also
provides a suite of confidence measures and graphical displays that allow the analyst to calculate the reliability
of the inferences (O‘Loughlin, 2000). While some disputes and concerns about King’s method still abound
(see, for example, Anselin, 2000), the EI (ecological inference) program, available from http://gking.harvard.edu,
that has been developed by King provides sufficient information to allow the analyst to decide how much
confidence he/she should place in the estimates.
A brief exegesis of the King ecological inference method is necessary to introduce its use in this
paper. In the absence of any aggregate census data for the constituencies or rayoni of Ukraine, I had to rely
on the data provided on the election commission. The quality of the ecological inferences relies heavily on
the quality of the data used to construct the inferences. Data sets with many geographical units, relatively low
heterogeneity, a proportional distribution across many categories, and a temporal coincidence in the
collection of the data sets all help to generate accurate inferences. With 675 cases, as in this study of Ukraine,
the estimates were reliable for most analyses, though the program failed in two instances (see Table 1). The
key predictor variable for the Kuchma vote was the level of turnout in each constituency. Did Kuchma
voters go to the polls to vote for their candidate at a significantly higher or lower rate than Symonenko
voters, and did these differences vary spatially across the country? To test the accuracy of the claims for the
presence of a regional factor in the country, regional differences in the turnout rates of the supporters of the
two candidates are measured using cartographic and statistical methods, with a null hypothesis of no
significant regional differences.
20
Table 1: EI estimates of the Turnout of Kuchma and Symonenko voters, Ukrainian Presidential runoff election 1998 Number
of Cases Average Turnout
Ratio
Kuchma Ratio
Estimated Turnout of Kuchma Voters
Estimated Turnout of Symonenko
Voters
Over (+) or Under (-)
Representation of Kuchma Voters
All Ukraine 675 .793 .516 .829 .655 +.036 Cities 196 .711 .548 .727 .660 +.016 Rayoni 489 .828 .511 .924 .721 +.096 West 156 .847 .795 .937 .538 +.090 Central 213 .778 .422 N/A N/A N/A South (inc. Crimea) 103 .707 .452 N/A N/A N/A East 203 .796 .452 .723 .786 -.072 N/A - EI estimates not provided because the estimates are not reliable. The distribution (density plot) of the estimates is very broad and the model fit is questionable. See King, 1997, Chapter 9 and pp. 242 ff.
In order to calculate the turnout rates for the Kuchma voters, we use the overall turnout rate (Bi) and the
Kuchma percentage (Wi) for each geographic unit, in this case, the 675 geographic units (rayoni and cities) of
Ukraine to calculate the overall rate of turnout for the Kuchma voters and the estimates for each of the
individual districts. The procedure follows that used by O’Loughlin (2000) to estimate the turnout of Nazi
party voters in the general election in the Weimar Republic in 1930. Using King’s notation, in the turnout
example, the independent variable, X, is the Kuchma runoff vote and the dependent variable is overall voter
turnout, T. An identity is used for combinations of the district values for Ti (turnout) and Xi (Kuchma
voters), Ti = ?ib Xi + ?iw (1 – Xi). The purpose of the EI modeling is to estimate ßb (the aggregate turnout
rate for Kuchma voters for the whole country) as well as the estimates for the individual rayoni and cities (675
units in all), ?ib . Combined with information about the bounds for each district, found by projecting the line
onto the horizontal (?ib , the Kuchma voter turnout) and the vertical (?iw, the non-Kuchma turnout) axes, the
EI method combines two earlier inference methods (King, 1996). Clearly, the narrower the bounds on the
axes, the stronger the chances of a plausible solution to what Anselin (2000) calls an “unobservable” value.
In the application of King’s EI methodology to the Ukrainian Presidential runoff data, five of the seven
21
analyses generated reliable estimates for the turnout rate of Kuchma voters but the distributional
requirements of the method precluded reliable estimates for the southern and central regions of Ukraine. (See
Table 1).
The second method used extensively in this paper also derives from the work of Gary King and his
associates (King, Tomz and Wittenberg, 2000). This methodology develops the presentation of statistical
results in a visual format. King and his co-authors argue correctly that too much of the impact and
importance of statistical analysis is hidden by the difficulty of interpretation of parameter values, coefficients
and significance tests in tabular form. In line with other initiatives in geography (Fotheringham, 1998) and
other social sciences (Cleveland, 1993; Tufte, 1983), the emphasis in this new approach is to take full
advantage of the large amount of information embedded in statistical analysis. It is especially the case that the
results of logit and other non-linear modeling exercises are hard to interpret when presented in the usual
tabular form and translation of the coefficients into graphical form helps the reader enormously in
Figure 3: Estimated Turnout of Kuchma Voters, Presidential Run-off Election
27
electoral geographic research should be to account for any underlying factors in the explanation of the voting
behavior so that the final map portrays an even distribution or no evidence of regional clustering of the
phenomenon of interest. However, King’s (1997, 25) own map of the ecological estimates of white turnout
in New Jersey elections shows clustering of high values near Newark and King speculates that a contextual
factor (though he does not use this term), proximity to the predominantly African-American city of Newark
might motivate higher involvement of whites in the electoral process. Similarly in Ukraine, it is possible that
regionally specific factors such as ethnic tension in Crimea between Tatars, Ukrainians and Russians might
influence the local turnout rates.
The dominant feature of the four maps of the ecological estimates of the Kuchma voter turnout is of
relatively even distributions, with a more modest concentration apparent in the west than was the case for
either the overall turnout or Kuchma vote percentage maps. Only the Crimean peninsula shows an oblast-
level concentration of values – in this case, of low values less than 81% turnout. All five districts of the port
city of Sevastopol and all but five of the other districts of the peninsula fall into this category. Overall,
Kuchma won just over half of the votes in Crimea but if his supporters had turned out here at the same rate
as they had elsewhere in the country, he would have added significantly to his strong majority. Exactly why
the Kuchma voter turnout in Crimea should be lowered is not evident; if the pattern identified elsewhere in
Ukraine holds in the peninsula, it is expected that Kuchma would be disproportionately supported by ethnic
Ukrainians. While the Kuchma voter turnout is low, it should also be noted that the overall turnout in
Crimea is low (Figure 1) and therefore, any regional advantage accruing to Symonenko was minimal.
As might be expected from studies of both electoral and grassroots political activism (Birch, 2000),
the west of Ukraine, especially Galicia, is the core of Kuchma support. Almost all his potential voters came
to the polls in this region (Figure 3). Political geographers have stressed the legacy of regional and local
historical memories and traditions in explaining the disparities in electoral maps, especially higher than
expected values (Agnew 1987). Galicia is such a region with a distinctive regional history as a result of
location in the Austrian-Hungarian empire (Bialasiewicz and O’Loughlin, 2001) and a pre- and post-
28
independence tradition of Ukrainian nationalist mobilization (Birch, 2000). The map of Kuchma voter
turnout provides further evidence of the importance of this distinctive regional legacy.
King’s (1997) ecological inference method has provided a major new tool for analysis in electoral
geography. While most of the attention since the method first came to prominence has been on national-
level estimates and checks against individual-level data to see if the method proves reliable, the value of the
unit-level estimates has largely gone unmarked, except by geographers. Though the reliability of the
disaggregated estimates is clearly less than the national-level figures, the values nonetheless are very useful
because they show geographic variation in a key relationship. Since individual level data are very rarely
available for historical study or for states in transition, researchers are often forced to use estimating methods
to arrive at individual level relationships. Though this was the motivation for the ecological inferential
techniques, the value of the district-level inferential methods should not be under-estimated. It can lead to a
renaissance of electoral geography because it enhances the possibility of examining traditional geographic
hypotheses about the relationship of a voter to his or her local context.
The Regional Geographies of Ukrainian Political Preferences
Underlying the voting choice of the 1999 Presidential election were the political preference structures of
Ukrainian citizens. While almost all Ukrainians (95% in 1999) are dissatisfied with the state of their
democracy and are worried about their declining standard of living, they are some significant regional, class,
ethnic and other differences apparent in public opinion polls (Ferguson, 1999). Unfortunately, these polls are
rarely consistent across the years, either in the nature of the questions asked, the distribution of the sample
across population groups, or geographic spread across all the regions. The latest large-scale reliable survey
data available from the Eurabarometer date from 1996. A comparison of results from the survey in the first
year, 1992, with the last year, 1996, allows some consideration of trends in the Ukrainian political setting and
whether preferences are underpinned by similar factors across the years.
29
Of the many questions asked of Ukrainians in the Eurobaromter polls, two questions stand out as
highly relevant to the state of politics in the country. The first question asked directly about the “Russian
dilemma”, phrasing the question as “As things stand now, with which of the following countries do you see
Ukraine’s future most closely tied up?” One of the options was Russia. The same question posed other
options, including preferences about a geopolitical orientation towards the Western states (U.S., European
Union and other European countries). The cleavages revealed in Ukraine on the Russia and Western
orientations are mirror images of each other. A second set of questions asked about the nature of
preferences for the economy and society - “Do you personally feel that the creation of a free market
economy, that is one largely free from state control, is right or wrong for Ukraine’s future?” Background
questions asked about age, gender, education, location, oblast residence, language use, ethnicity, urban or
rural residence. The questions about income and standard of living were amplified by a key question about
recent changes in the family situation, phrased as: “Compared to 12 months ago, do you think the financial
situation of your household has… .”; options included got a lot better, got a little better, stayed the same, got a
little worse, got a lot worse, and don’t know. In the analyses reported here, this change in the standard of
living is a key explanation of preferences for the western model (orientation to the West) and the economic
model (free market or socialism).
To illustrate the point made repeatedly in this paper that the strength of the regional factor in
Ukraine is correlated with the nature of the issue, the results of a logit regression for the “satisfaction with
democracy” question are presented in Table 2 for 1992 and 1996. The model yields an explanation that those
with a Ukrainian mother tongue, males, and households with an improvement in standard of living in the
previous year were more satisfied with the state of Ukrainian democracy in 1992. Those with a higher
income were more dissatisfied but importantly, the regional factor was not significant. By 1996, Ukrainian-
speakers and households with improved standard of living were still more satisfied, as were older citizens, but
better educated respondents were dissatisfied. The regional variable (using a four categorical variable for
respondents from the west, center, east and south plus Crimea) did not enter the equation at any level of
significance and is thus excluded from consideration. The results are not too surprising since the
30
Table 2: Logistic Regression Estimates for Satisfaction with the State of Democracy in Ukraine 1992 and 1996 Satisfaction with Democracy 1992* Satisfaction with Democracy 1996**
Predictor Coefficient Std. Error z Coefficient Std. Err. z Household Finance .397 .060 6.608 .496 .074 6.659
Mother Tongue .486 .155 3.132 .746 .181 4.114 Income -.004 .002 -1.961 -- -- -- Region -.025 .072 -0.346 -- -- -- Male -.262 .142 -1.845 -0.14 .076 -0.194 Age -- -- -- .011 .005 2.285
Non-significant z –values at a = .10 are underlined * Observations = 1400; ? 2 = 59.16 (significant at a = .001) ** Observations = 1200; ? 2 = 88.09 (significant at a = .001)
dissatisfaction with the nature of Ukrainian democracy was pervasive across all populations and regions, and
when some key personal characteristics are controlled, the regional factor disappears. Geographers do not
always expect to find regional significance but if we start with an expectation that compositional
characteristics provide adequate explanation of preferences and behavior, we can reject that hypothesis when
careful statistical analysis suggests the persistence of the regional and local elements as part of the
explanations.
A more elegant visual presentation and a more easily interpreted display of the results of the logit
models can be seen in Figure 4. Using the point estimates and the parameters from a logit model, the
simulation used a large sample (usually 1000 draws) to draw a value of Y (the dependent variable) conditional
on one chosen value of each explanatory variable. The expected or mean value of Y is computed in this
manner as well as measures of uncertainty around the mean. Values of the independent variables in the
equation can be set to fixed values (say, age set to 30 and gender to men) and then, the CLARIFY program
will generate quantities of interests (King, Tomz and Wittenberg, 2000). We can, for example, then compare
the expected level of satisfaction in Ukraine of men aged 30 to women aged 30 or to men aged 40, thus
computing the difference in the dependent variable from changing gender, holding age constant, or a 10 year
increase in age, holding gender constant. Generation of many such expected outcomes holding various
Age in 5 Year Intervals
0
0.2
0.4
0.6
0.8
20 25 30 35 40 45 50 55 60 65 70 75 80 85
Figure 4: Probabilities of Support for the Western Model, 1992 and 1996
Est
imat
ed P
roba
bilit
yE
stim
ated
Pro
babi
lity
0
0.2
0.4
0.6
0.8
Support for Western Model 1992
20 25 30 35 40 45 50 55 60 65 70 75 80 85Age in 5 Year Intervals
Support for Western Model 1996
Mean Value with Confidence Interval
E. non-Ukrainians W. Ukrainians E. Ukrainians
Mean Value with Confidence Interval
E. non-Ukrainians W. Ukrainians E. Ukrainians
John O'Loughlin
John O'Loughlin
31
combinations of independent predictors constant allows presentation in a graphical form. In Figures 4-11, I
highlight the regional element in the displays because that factor is the focus of my study.
A simple display of the expected outcomes of support for the Western model (future should be ties
to either the U.S., European Union or other European state) is shown for both the 1992 and 1996
Eurobarometer surveys in Ukraine in Figure 4. The logit model of orientation to the West contained
predictors for region, age, ethnicity, education, income, and change in the standard of living over the previous
year. All independent variables, except the regional factor, ethnicity and age are set at their median values to
approximate an “average” respondent in these categories and a focus on the three factors of region, ethnicity
and age. The graphs in Figure 4 display both the expected mean values and the five percent confidence
intervals. Ethnic Ukrainians in the west have a significantly higher level of support for the western model
than the other two groups (Ukrainians in the east and Russians in the east) at all age intervals. Not only is the
mean value higher but also the confidence intervals do not intersect with those of the other samples. The gap
between the groups widens perceptibly between 1992 and 1996 and as is common throughout the former
Communist states, support for the Western model is strongest for the younger populations and decreases
gradually for all groups. There is no significant difference between the ethnic groups in the east of Ukraine,
with non-Ukrainians (almost all are Russians) and Ukrainians showing similar mean values in 1992; by 1996,
the gap between these groups in the eastern region has grown as a result of more support for the Western
model by the Ukrainian sample but the confidence intervals overlap so that the gap is not statistically
significant. A verbal interpretation of the 1992 graph might be that a typical 50 year old Ukrainian in the west
of the country have a 45% chance of choosing a Western country as a partner, while the same person in the
east has only a 22% chance of picking this option for the future. This difference is statistically significant. As
a method to clarify regional differences, this methodology offers a significant advantage over tabular formats.
The remaining seven analyses are presented in a box-plot format. Box-plots offer an advantage over
graphs because the distributional characteristics of the estimates are more clearly displayed. On each of the
figures 5-11, the median value of the estimated probabilities is shown, with below average values shaded. On
a box-plot, 50 percent of the cases have values within the box (from the 25th to the 75th percentile) and the
32
whiskers extending from the box mark the largest and smallest observed values that are not outliers. Extreme
points beyond the whiskers are identified individually. In comparing the plots, it is important to note that the
scale of the Y-axis is not standardized across the analyses while each of the sub-populations is denominated
individually on the horizontal line. Logit models with both regional and non-regional controls are presented
but it should be stressed again that not all the analyses showed significant regional conditions after the
compositional effects are controlled.
Satisfaction with democracy estimates are shown for 1992 and 1996 in Figures 5 and 6, and the logit
models from which the estimated probabilities are derived are shown in Table 2. Across all groups,
satisfaction with the state of Ukrainian democracy is low, but there are some significant compositional effects
as seen in the box-plots. The biggest difference is between the language groups, Russophones and
Ukrainophones, while the regional differences across these groups are almost non-existent at less than .01
(Figures 5 and 6). While a 6 point difference (1992) or 14 point difference (1996) does not seem dramatic, it
is significant given the generally low level of satisfaction with democracy. Most significantly, the gap between
the language groups widened between 1992 and 1996, though the controls were slightly different. It is still
too early to speculate about the fact that inter-group differences are larger than regional differences; generally,
the most dangerous scenario for a society is when an ethnic-linguistic minority is regionally-concentrated in a
periphery and subject to economic and cultural discrimination (Gurr and Moore, 1997). Ukraine does not fit
this scenario and there are enough cross-regional and cross-cultural linkages, as well as careful state policy, to
head off any murmurings of ethno-territorial mobilization.
The clearest expressions of the regional factor in the logit analyses surround the question of whether
Ukraine’s future lies with Russia. In 1992, the regional factor (residence in one of four regions) is highly
significant, though by 1996, this factor had ebbed to insignificance to be replaced by ethnicity, occupation
and the trend in household finances as statistical predictors of preferred relations with Russia (Table 3).
Only language group and big-city residence remains as significant predictors across the samples. The
estimated probabilities for different combinations of sub-groups are shown on Figures 7-9. Combinations of
the ethnic groups, language groups and the four regions are presented for 1996 in Figure 7 and arranged for
10/2/00 JohnO/boxplots_10/2/00/B
_box_uksd_rusd.fh8
RussianSpeakers
Center
RussianSpeakers
East
RussianSpeakers
West
RussianSpeakers
South
.11 .11.10 .10
Ukrainian��������Center
Ukrainian��������
East
UkrainianSpeakers
South
Ukrainian��������
West
Est
imat
ed P
roba
bilit
y of
Sat
isfa
ctio
n w
ith D
emoc
racy
199
2
.3
0
.225
.15
.075
.16 .16 .16 .17
John O'Loughlin
Figure 5: Estimated Probability of Satisfaction with Democracy Ukraine 1992 by Region and Language
10/2/00 JohnO/boxplots_10/2/00boxpic3_4.fh8
Ukrainian speakers
West
Russianspeakers
West
Ukrainianspeakers
South
Russian speakers
South
.35
.20
.35
.19
0
.60
.45
.30
.15
.34
.19
.34
.20
UkrainianspeakersCenter
Ukrainianspeakers
East
Russianspeakers
East
Est
imat
ed, P
roba
bilit
y of
Sat
isfa
ctio
n w
ith D
emoc
racy
, 199
6
RussianspeakersCenter
John O'Loughlin
Figure 6: Estimated Probability of Satisfaction with Democracy Ukraine 1996 - by region and language
33
Table 3: Logistic Regression Estimates for Views on a Future tied to Russia, Ukraine 1992 and 1996 Future tied to Russia 1992* Future tied to Russia 1996**
Predictor Coefficient Std. Error z Coefficient Std. Err. z Religion -.292 .127 -2.299 -- -- --